Nearly 3 in 4 enterprises hear customer complaints on AI personalization

AI-powered personalization now decides what most customers see first, from the products on your homepage to the offers at checkout. Spending on it keeps climbing, and customers are often the ones catching the mistakes.

Among enterprises that hit an error in the past 12 months, 75% found out through customer complaints. These errors create for your users, cost your business resources to fix, and undercut the customer experience you were trying to build.

Here's a look at what's going wrong and how you can fix it.

78% of enterprises see measurable results from AI personalization

Enterprises are spending more on AI personalization every year, with 89% increasing their investment over the past two years. This commitment is paying off, with 78% reporting measurable business results and nearly as many reporting improved customer experience.

Those results depend on correctly interpreting behavioral signals. A customer who bought running shoes last week should be shown socks and other accessories, not an ad for more shoes or a set of golf clubs. 

Getting personalization wrong undercuts the trust your investment was meant to build. And these misses happen frequently, with most teams hearing about them from the customers who ran into them.

75% still hear customer complaints, and 55% see errors weekly

Personalization errors are common across enterprises, with 87% hitting at least one in the past 12 months. Most come down to the system acting on an outdated picture of the customer, recommending something they've already bought, repeating an offer they've already dismissed, or pushing an item that's been out of stock for weeks.

Among enterprises that had encountered one of these mistakes, 75% discovered it through customer complaints or negative feedback. By then the issue already reached someone and bothered them enough to respond.

These errors also repeat, with 55% of enterprises seeing them at least weekly. The teams that catch them early are the ones watching customer behavior in real time, which gives them a chance to fix the experience before anyone has a reason to complain.

Personalization is often really good at picking the right product. It’s not always good at picking the right moment. When it misses, you can see it in the session. Someone’s halfway through checkout, gets a carousel of new stuff to look at, and wanders off. Or they get a recommendation for the jacket they bought last week and scroll past. They still want to buy something! The personalization just stopped helping.
- Liz Russell, Manager, Professional Services

Only 22% catch personalization errors in their own behavioral data

Teams already use behavioral data to power personalization, more than purchase history, declared preferences, or third-party inputs. But watching that same data for mistakes is a different story. Only 22% catch an error that way, while 34% wait on a system flag and 32% hear it from a customer.

Finding these takes more than spotting a bad recommendation. To know that running shoes on the homepage are a mistake, you need the purchase from last week, the session where the customer scrolled past that same product twice, and what they did next.

One reason customers find errors before you do is shiny object syndrome. Teams watch a launch closely for a week or two, then move on to the next thing. But personalization errors don’t show up at launch. They show up months later, when inventory changes or purchase data doesn’t sync. These also aren’t bugs QA will catch. Nothing is technically broken. The system returned a valid recommendation; it was just wrong for that person at that moment. You can’t test every customer against every product, so the customer ends up finding it.
- Liz Russell, Manager, Professional Services

Only 28% of enterprises personalize based on what a customer does during the current session. Working from the current session is what makes a mistake visible while it's still happening.

63% believe personalization errors cost them revenue, but only 37% have measured it

Letting these errors go unchecked has a revenue cost, with 63% of enterprises saying personalization mistakes have cost them sales in the past 12 months. Thirty percent put their annual losses at $500,000 or more, and 13% say the figure runs past $1 million.

Only about a third have traced those losses back to the errors that caused them. Doing that means following what the customer did after a bad recommendation, whether they abandoned the session, skipped the purchase, or never came back.

When you can't put a number on a problem, it's hard to justify fixing it. Behavioral data gives you that number, along with supporting metrics like and insight into what the customer was doing when it happened.

Here’s how an enterprise company cut its error rates by 67% with behavioral insights

MOO is a print and branded merchandise company with over 450 employees across the UK and the United States, and a useful example of AI ecommerce personalization running at scale.

In early 2024, MOO launched a self-serve experience for branded merchandise. Customers were dropping off before converting, and the team couldn't tell why. Google Analytics showed them where people left, but not what happened in those sessions.

Fullstory surfaced the rage clicks and dead clicks that had gone unnoticed, and tied each to the stage of the journey where it occurred. Using this data, MOO cut error rates on the beta product page by 67%, dropped checkout abandonment by 12%, and increased the share of customers viewing two or more products by 12%.

Improve your AI personalization with Fullstory

The clearest way to catch these errors is to look at what customers do when personalization misses the mark. Their behavior reveals a problem even when the system itself shows no error.

Fullstory makes this possible by indexing every interaction through Fullcapture, and StoryAI turns that data into recommendations for what to fix first. Unlike AI-powered personalization solutions that only act on data, this shows you what you missed.

That gives teams a chance to catch personalization errors before customers do and improve conversion rates.

Catch personalization errors before your customers do.

Get a demo ➜

Methodology

This survey was conducted by Centiment for Fullstory. Results are based on 573 completed surveys from U.S. professionals in Product Management, Engineering, Data & Analytics, Customer Experience/Support, and IT/Security roles, spanning company sizes and industries including retail, financial services, travel and hospitality, SaaS, and media and entertainment.